摘要
肺结节CT图像的相似性检索中使用的特征通常依赖于手工提取的像素级别的图像特征的准确性与完整性,造成检索匹配精度低、检索速度慢等问题。为解决上述问题,提出一种基于深度哈希的肺结节CT图像检索方法,使用深度学习强大的数据处理能力对肺结节的语义级别特征进行深度提取,有效结合哈希算法,实现检索过程由粗到精的有效操作,返回最为相似的肺结节图像。实验结果表明了所提方法的有效性。
The characteristics used in the similarity search of pulmonary nodule CT images usually depend on the accuracy and completeness of the image features extracted at the pixel level manually,resulting in problems such as low retrieval accuracy and low retrieval speed.To solve the above problems,a retrieval method of pulmonary nodules CT images based on deep hashing was proposed.The powerful data processing capabilities of deep learning were used to extract the semantic level features of pulmonary nodules,and the hash algorithm was effectively combined,the retrieval process was operated from coarse to fine and the most similar pulmonary nodule image was returned.Performances of the proposed method were demonstrated by the experiment.
作者
宋云霞
强彦
唐笑先
SONG Yun-xia 1,QIANG Yan 1 ,TANG Xiao-xian 2(1.College of Computer Science and Technology,Taiyuan University of Technology,Taiyuan 030024,China;2.CT Room,Shanxi Provincial People ’s Hospital,Taiyuan 030012,Chin)
出处
《计算机工程与设计》
北大核心
2018年第7期1954-1959,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(61373100)
虚拟现实技术与系统国家重点实验室开放基金项目(BUAA-VR-17KF-14
BUAA-VR-17KF-15)
山西省回国留学人员科研基金项目(2016-038)
关键词
肺结节图像
卷积神经网络
有监督哈希
图像检索
主成分分析
pulmonary nodule images
convolutional neural networks
supervised hashing
image retrieval
principal components analysis